am f7322b16: Make getNextWordAndNextToken output code point count.

* commit 'f7322b166b88f72b19509d8416700d4ec8ea7753':
  Make getNextWordAndNextToken output code point count.
main
Keisuke Kuroyanagi 2014-06-23 10:26:40 +00:00 committed by Android Git Automerger
commit 179eb125c4
10 changed files with 45 additions and 47 deletions

View File

@ -304,17 +304,18 @@ static jint latinime_BinaryDictionary_getNextWord(JNIEnv *env, jclass clazz,
jlong dict, jint token, jintArray outCodePoints) {
Dictionary *dictionary = reinterpret_cast<Dictionary *>(dict);
if (!dictionary) return 0;
const jsize outCodePointsLength = env->GetArrayLength(outCodePoints);
if (outCodePointsLength != MAX_WORD_LENGTH) {
AKLOGE("Invalid outCodePointsLength: %d", outCodePointsLength);
const jsize codePointBufSize = env->GetArrayLength(outCodePoints);
if (codePointBufSize != MAX_WORD_LENGTH) {
AKLOGE("Invalid outCodePointsLength: %d", codePointBufSize);
ASSERT(false);
return 0;
}
int wordCodePoints[outCodePointsLength];
memset(wordCodePoints, 0, sizeof(wordCodePoints));
const int nextToken = dictionary->getNextWordAndNextToken(token, wordCodePoints);
int wordCodePoints[codePointBufSize];
int wordCodePointCount = 0;
const int nextToken = dictionary->getNextWordAndNextToken(token, wordCodePoints,
&wordCodePointCount);
JniDataUtils::outputCodePoints(env, outCodePoints, 0 /* start */,
MAX_WORD_LENGTH /* maxLength */, wordCodePoints, outCodePointsLength,
MAX_WORD_LENGTH /* maxLength */, wordCodePoints, wordCodePointCount,
false /* needsNullTermination */);
return nextToken;
}
@ -555,12 +556,13 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
// TODO: Migrate historical information.
int wordCodePoints[MAX_WORD_LENGTH];
int wordCodePointCount = 0;
int token = 0;
// Add unigrams.
do {
token = dictionary->getNextWordAndNextToken(token, wordCodePoints);
const int wordLength = CharUtils::getCodePointCount(MAX_WORD_LENGTH, wordCodePoints);
const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints, wordLength);
token = dictionary->getNextWordAndNextToken(token, wordCodePoints, &wordCodePointCount);
const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints,
wordCodePointCount);
if (dictionaryStructureWithBufferPolicy->needsToRunGC(true /* mindsBlockByGC */)) {
dictionaryStructureWithBufferPolicy = runGCAndGetNewStructurePolicy(
std::move(dictionaryStructureWithBufferPolicy), dictFilePathChars);
@ -569,8 +571,8 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
return false;
}
}
if (!dictionaryStructureWithBufferPolicy->addUnigramEntry(wordCodePoints, wordLength,
wordProperty.getUnigramProperty())) {
if (!dictionaryStructureWithBufferPolicy->addUnigramEntry(wordCodePoints,
wordCodePointCount, wordProperty.getUnigramProperty())) {
LogUtils::logToJava(env, "Cannot add unigram to the new dict.");
return false;
}
@ -578,9 +580,9 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
// Add bigrams.
do {
token = dictionary->getNextWordAndNextToken(token, wordCodePoints);
const int wordLength = CharUtils::getCodePointCount(MAX_WORD_LENGTH, wordCodePoints);
const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints, wordLength);
token = dictionary->getNextWordAndNextToken(token, wordCodePoints, &wordCodePointCount);
const WordProperty wordProperty = dictionary->getWordProperty(wordCodePoints,
wordCodePointCount);
if (dictionaryStructureWithBufferPolicy->needsToRunGC(true /* mindsBlockByGC */)) {
dictionaryStructureWithBufferPolicy = runGCAndGetNewStructurePolicy(
std::move(dictionaryStructureWithBufferPolicy), dictFilePathChars);
@ -589,8 +591,8 @@ static bool latinime_BinaryDictionary_migrateNative(JNIEnv *env, jclass clazz, j
return false;
}
}
const PrevWordsInfo prevWordsInfo(wordCodePoints, wordLength,
false /* isStartOfSentence */);
const PrevWordsInfo prevWordsInfo(wordCodePoints, wordCodePointCount,
false /* isBeginningOfSentence */);
for (const BigramProperty &bigramProperty : *wordProperty.getBigramProperties()) {
if (!dictionaryStructureWithBufferPolicy->addNgramEntry(&prevWordsInfo,
&bigramProperty)) {

View File

@ -145,10 +145,11 @@ const WordProperty Dictionary::getWordProperty(const int *const codePoints,
codePoints, codePointCount);
}
int Dictionary::getNextWordAndNextToken(const int token, int *const outCodePoints) {
int Dictionary::getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount) {
TimeKeeper::setCurrentTime();
return mDictionaryStructureWithBufferPolicy->getNextWordAndNextToken(
token, outCodePoints);
token, outCodePoints, outCodePointCount);
}
void Dictionary::logDictionaryInfo(JNIEnv *const env) const {

View File

@ -103,7 +103,8 @@ class Dictionary {
// Method to iterate all words in the dictionary.
// The returned token has to be used to get the next word. If token is 0, this method newly
// starts iterating the dictionary.
int getNextWordAndNextToken(const int token, int *const outCodePoints);
int getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount);
const DictionaryStructureWithBufferPolicy *getDictionaryStructurePolicy() const {
return mDictionaryStructureWithBufferPolicy.get();

View File

@ -104,7 +104,8 @@ class DictionaryStructureWithBufferPolicy {
// Method to iterate all words in the dictionary.
// The returned token has to be used to get the next word. If token is 0, this method newly
// starts iterating the dictionary.
virtual int getNextWordAndNextToken(const int token, int *const outCodePoints) = 0;
virtual int getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount) = 0;
virtual bool isCorrupted() const = 0;

View File

@ -478,10 +478,9 @@ const WordProperty Ver4PatriciaTriePolicy::getWordProperty(const int *const code
return WordProperty(&codePointVector, &unigramProperty, &bigrams);
}
int Ver4PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outCodePoints) {
// TODO: Return code point count like other methods.
// Null termination.
outCodePoints[0] = 0;
int Ver4PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount) {
*outCodePointCount = 0;
if (token == 0) {
mTerminalPtNodePositionsForIteratingWords.clear();
DynamicPtReadingHelper::TraversePolicyToGetAllTerminalPtNodePositions traversePolicy(
@ -498,13 +497,8 @@ int Ver4PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const
}
const int terminalPtNodePos = mTerminalPtNodePositionsForIteratingWords[token];
int unigramProbability = NOT_A_PROBABILITY;
const int codePointCount = getCodePointsAndProbabilityAndReturnCodePointCount(
*outCodePointCount = getCodePointsAndProbabilityAndReturnCodePointCount(
terminalPtNodePos, MAX_WORD_LENGTH, outCodePoints, &unigramProbability);
if (codePointCount < MAX_WORD_LENGTH) {
// Null termination. outCodePoints have to be null terminated or contain MAX_WORD_LENGTH
// code points.
outCodePoints[codePointCount] = 0;
}
const int nextToken = token + 1;
if (nextToken >= terminalPtNodePositionsVectorSize) {
// All words have been iterated.

View File

@ -134,7 +134,8 @@ class Ver4PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy {
const WordProperty getWordProperty(const int *const codePoints,
const int codePointCount) const;
int getNextWordAndNextToken(const int token, int *const outCodePoints);
int getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount);
bool isCorrupted() const {
return mIsCorrupted;

View File

@ -391,7 +391,9 @@ const WordProperty PatriciaTriePolicy::getWordProperty(const int *const codePoin
return WordProperty(&codePointVector, &unigramProperty, &bigrams);
}
int PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outCodePoints) {
int PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount) {
*outCodePointCount = 0;
if (token == 0) {
// Start iterating the dictionary.
mTerminalPtNodePositionsForIteratingWords.clear();
@ -409,8 +411,8 @@ int PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outC
}
const int terminalPtNodePos = mTerminalPtNodePositionsForIteratingWords[token];
int unigramProbability = NOT_A_PROBABILITY;
getCodePointsAndProbabilityAndReturnCodePointCount(terminalPtNodePos, MAX_WORD_LENGTH,
outCodePoints, &unigramProbability);
*outCodePointCount = getCodePointsAndProbabilityAndReturnCodePointCount(terminalPtNodePos,
MAX_WORD_LENGTH, outCodePoints, &unigramProbability);
const int nextToken = token + 1;
if (nextToken >= terminalPtNodePositionsVectorSize) {
// All words have been iterated.

View File

@ -137,7 +137,8 @@ class PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy {
const WordProperty getWordProperty(const int *const codePoints,
const int codePointCount) const;
int getNextWordAndNextToken(const int token, int *const outCodePoints);
int getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount);
bool isCorrupted() const {
return mIsCorrupted;

View File

@ -489,10 +489,9 @@ const WordProperty Ver4PatriciaTriePolicy::getWordProperty(const int *const code
return WordProperty(&codePointVector, &unigramProperty, &bigrams);
}
int Ver4PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outCodePoints) {
// TODO: Return code point count like other methods.
// Null termination.
outCodePoints[0] = 0;
int Ver4PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount) {
*outCodePointCount = 0;
if (token == 0) {
mTerminalPtNodePositionsForIteratingWords.clear();
DynamicPtReadingHelper::TraversePolicyToGetAllTerminalPtNodePositions traversePolicy(
@ -509,13 +508,8 @@ int Ver4PatriciaTriePolicy::getNextWordAndNextToken(const int token, int *const
}
const int terminalPtNodePos = mTerminalPtNodePositionsForIteratingWords[token];
int unigramProbability = NOT_A_PROBABILITY;
const int codePointCount = getCodePointsAndProbabilityAndReturnCodePointCount(
*outCodePointCount = getCodePointsAndProbabilityAndReturnCodePointCount(
terminalPtNodePos, MAX_WORD_LENGTH, outCodePoints, &unigramProbability);
if (codePointCount < MAX_WORD_LENGTH) {
// Null termination. outCodePoints have to be null terminated or contain MAX_WORD_LENGTH
// code points.
outCodePoints[codePointCount] = 0;
}
const int nextToken = token + 1;
if (nextToken >= terminalPtNodePositionsVectorSize) {
// All words have been iterated.

View File

@ -113,7 +113,8 @@ class Ver4PatriciaTriePolicy : public DictionaryStructureWithBufferPolicy {
const WordProperty getWordProperty(const int *const codePoints,
const int codePointCount) const;
int getNextWordAndNextToken(const int token, int *const outCodePoints);
int getNextWordAndNextToken(const int token, int *const outCodePoints,
int *const outCodePointCount);
bool isCorrupted() const {
return mIsCorrupted;